a–d, Western Africa (a), central Africa (b), southeast Asia (c) and the Guiana Shield (d). Corridors are shown in white, protected areas in semi-transparent grey and carbon density of woody vegetation as a gradient from low density in red to high density in green.
We acquired protected area boundaries from the 2010 release of the WDPA (ref. 27). We selected designated protected areas for all International Union for Conservation of Nature categories to serve as anchors for corridors (Supplementary Methods).
As our goal was to map corridors traversing high VCS between protected areas, we calculated a landscape resistance surface so that high-VCS areas would be less costly to traverse and low-VCS areas would be more costly to traverse. See ref. 6 for details on creation of VCS maps. We calculated the resistance surface by taking the inverse of VCS values and recoding wide water bodies with a high-resistance value (Supplementary Methods).
For each terrestrial protected area in the WDPA we used Thiessen polygons to define the set of first-order neighbours using landscape resistance as a measure of separation between protected area boundaries 28 (Supplementary Fig. 1) and then mapped least-cost corridors between each pair of protected areas using the conditional minimum transit cost algorithm29 (Supplementary Methods). As a sensitivity analysis, we constructed corridors using another publicly available, pan-tropical biomass data set15.
We compared the efficiency of a corridor approach with a BAU VCS preservation approach where, within a given country, spatial location or contiguity of pixels was not considered, only biomass. To do this, we identified the minimum set of pixels that, when summed, equal the amount of VCS within corridors. We then subtracted the BAU area from corridor area and calculated the per cent difference, relative to corridor area, in the area needed to preserve the same amount of carbon as is found in corridors.
Threat of deforestation in corridors and BAU areas was estimated across the tropics using the human footprint data set13 and for the Amazon using spatially explicit deforestation projections out to the year 2030 (ref. 14). We resampled the human footprint data to match the resolution of the VCS grids and summarized human footprint values in corridors and BAU areas. For the Amazon, we resampled deforestation projections to match the VCS grids and then calculated the fraction of each corridor projected to be deforested from 2002 to 2030.
We used the Global Administrative Areas database (http://www.gadm.org/) to summarize corridor biomass by country. We calculated mean VCS density by country and then multiplied this by the area of each country to arrive at an estimate of VCS in each country. We then repeated those steps after intersecting a binary representation of the corridor map with the VCS map and then masking out protected areas, thereby calculating unprotected VCS in corridors by country.
For the case study of the Legal Amazon, we used spatially explicit, 2-km resolution, modelled opportunity costs for soy, cattle and timber10 to estimate costs of foregone rents associated with corridor protection. For each pixel, we calculated the maximum net present value of potential land uses assuming a high-opportunity cost scenario. The calculations are as follows:
where NPV soy, NPV cattleandNPV timber are net present value of returns per hectare from soybean farming, cattle ranching and timber harvesting, respectively, and OC is opportunity cost in dollars per hectare.
We used two measures of biodiversity, species richness and a weighting of species richness by range size, termed endemism richness30. We downloaded extent of occurrence records for all terrestrial mammals globally 31. We gridded these geographic information system coverages at ~500 m resolution and calculated two measures of biodiversity: species richness, calculated by summing the number of ranges intersecting a given pixel; and endemism richness, calculated as follows:
where P is the total number of pixels in a species range, S is the number of species ranges that cover a given pixel and ER is the sum of inverse range fractions that cover a given pixel.
For the TOPSIS analysis (Supplementary Methods), we summarized species richness, endemic species richness, VCS and deforestation threat within corridors using ArcGIS zonal statistics. For each corridor we calculated fraction of corridor projected to be deforested, mean VCS and maximum values for the richness variables. We then calculated TOPSIS scores and divided them by EOC to rank corridor suitability in terms of average cost for the given criteria. We calculated VCS, biodiversity and deforestation threat as positive criteria to identify the most threatened corridors with high VCS and biodiversity values.